Every other catalogue counts. agentmods measures.
Skills, subagents, commands, hooks, Cursor rules and MCP servers for Claude Code, Cursor, Codex and Copilot, measured.
Each page says what the mod costs you in tokens on every session, what a security scan found and why, whether it is the original or one of the copies, and what changed since last week.
npx agentmods cost tells you what your current setup costs every session
Every mod is filed by what it actually does — read and judged one at a time, not matched on keywords. Each category opens the full list; the Top menu holds the measured hundred.
They are all the same shape of thing: a file you drop in that changes how the agent behaves. Splitting them across nine sites is what everyone else does, and it is why nobody can tell you what your whole stack costs.
Loaded on demand. Two digits per session until invoked.
24,473Subagents with their own prompt. Injected whole.
27,606Slash commands, one file per job.
55,644Servers. The cost is their tool schemas, not the config.
23,936Cursor rules, loaded whole into a matching conversation.
198AGENTS.md, CLAUDE.md, copilot-instructions.
1,185Shell commands your agent runs. Read before installing.
90Configuration, so you can see what others turn on.
22,548Bundles of the above, shipped together.
The six mods whose repositories gained the most stars in the last seven days — movement, not accumulated fame.
Skill Claude CodeCodex
Reverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.
Skill Codex
Fetch, inspect, crawl, and capture live, JavaScript-rendered websites with Moli. Use when Codex needs current web content, web research, fact lookup, link following, a bounded crawl, client-rendered or response-gated content, network diagnostics, or a standalone HTML, Markdown, JSON, semantic-tree, viewport or…
Cursor rule Cursor
Ponytail, lazy senior dev mode. Always pick the simplest solution that works.
Skill Claude CodeCodex
An internet-search and platform-access router for finding information across websites and services such as Reddit, X, GitHub, YouTube, and job sites.
Plugin Claude Code
Bundles 38 skills · 1,293 tokens together
Matt Pocock's agent skills for real engineering: grilling, spec/ticket flows, TDD, code review, domain modelling and more. Plug-and-play, not vibe coding.
Plugin Claude Code
Bundles 1 skill, 1 command, 4 agents, 5 hooks · 508 tokens together
Design fluency for frontend development. 1 skill with 24 commands (/impeccable polish, /impeccable audit, /impeccable critique, etc.) and curated anti-pattern detection.
Mods people are picking up this week, sorted by what they add to every session before you have asked anything. No other catalogue can offer this sort, because no other catalogue counts.
Skill Claude CodeCodex
Glance at one fresh headline and drop it into a single sentence in your own words — a casual, half-distracted aside between songs, never an anchor or newsreader voice, never the headline read verbatim, no editorialising, no "in other news". Steer to the lighter, cultural, or human-interest item; if everything on the…
Skill Claude CodeCodex
Clinical study design, statistical analysis, and regulatory compliance for medical research.
Skill Claude CodeCodex
Allow a viewer to subscribe to this agent for recurring USDC payments. The agent charges 5 USDC per week automatically — no re-signing needed — until the viewer revokes or the delegation expires (90 days).
Skill Claude CodeCodex
Use gh as the primary interface for GitHub data and actions. Keep this skill focused on basic operations and command safety, not maintainer triage strategy.
Skill Claude CodeCodex
You have a newsfetch tool that fetches raw news headlines and article content from multiple sources (Google News RSS, Hacker News API, Yahoo News, Substack, Medium). Use it to build daily news digests for the user.
Command Claude Code
Strictly follow the project constitution.
Every catalogue shows that a mod exists and how many stars sit behind it. None of them tells you what it costs, whether it is safe, whether it is the original, or whether it changed under you.
Per session and once invoked, counted locally with o200k_base and priced per model. A stack of six agents is twenty thousand tokens before you have typed anything.
Hidden Unicode, instruction-override phrasing, curl piped into a shell, recursive deletes, credential paths. A grade from A to F, and the exact line behind every finding.
Bodies are simhashed and clustered. The one with the most behind it is the original; the rest say how identical they are and point at it instead of competing with it.
Hashed on every crawl, with the history kept rather than the latest state alone. A supply-chain change to an agent config is a question of when, not whether.
Stars belong to the repository, not the mod: one repo with 170k stars and 139 skills would own the first three pages. Reputation enters through log10 and carries a quarter of the weight.
Every mod stays where its author put it and every page links to the source file. Bodies appear only under a permissive licence we can name; no LICENSE means all rights reserved, an unidentifiable one means unknown terms, and those pages say which.
What a mod is, what the numbers mean, and what this catalogue does and does not do.
agentmods, one word, at agentmods.dev, is a catalogue of mods for coding agents: skills, subagents, commands, hooks, rules, instructions files, MCP servers and plugins for Claude Code, Cursor, Codex and Copilot. Every entry is measured rather than merely listed: what it costs in tokens per session, what a security scan found and why, whether it is the original or a copy, and what changed since last week. The same data is on the command line as npx agentmods.
Anything you add to a coding agent to change how it behaves: a skill, a subagent, a slash command, a hook, a Cursor rule, an instructions file, an MCP server, a plugin bundle. They are all the same shape of thing — a file you drop in that the agent then reads — so they belong in one catalogue rather than nine.
Because a mod is not free to have installed. An agent or a Cursor rule is injected into the context of every session whether or not you use it, and a stack of them can be twenty thousand tokens before you have asked anything. Every page here prints what the mod costs per session and what it costs once invoked, counted locally with the o200k_base tokenizer.
It is a static scan, not an audit, and the page says so. It looks for hidden Unicode, instruction-override phrasing, shell pipes into a downloaded script, recursive deletes, privilege escalation, credential paths and encoded payloads. Every finding is printed with the line that produced it, so you can judge whether it matters in context rather than trusting a letter.
Because most of them are. Bodies are simhashed and near-identical documents are clustered; the one with the most behind it is treated as the original and the rest point at it. A catalogue that lists nine copies of the same skill as nine results is making the search harder, not easier.
No. Every mod stays where its author put it, and every page links to the source file. Bodies are shown only where the repository carries a permissive licence we can name. No LICENSE means all rights reserved, and a LICENSE we cannot identify means terms we will not guess at; either way those pages carry the metadata, the measurements and a link.
611,395 mods, measured four ways. Search by name, or sort a whole category by what it adds to your context.